kennwerte-db

Open-source construction cost benchmark database for Swiss public buildings.
Collects, structures, and presents cost Kennwerte (CHF/m² GF, CHF/m³ GV, BKP/eBKP-H breakdowns)
from realised projects to support early-stage cost estimation and portfolio-level cost analysis.

Live Demo

Features

• Gallery, List, Map, Dashboard — four views for browsing and comparing construction projects
• Detail view — SIA 416 volumes/areas, eBKP-H and BKP cost breakdowns, peer comparison box plots, image gallery with lightbox
• Scatter plot — GF vs. CHF/m² GF across the full dataset, colored by construction type
• Filters — by source, category, canton, construction type, country, data quality, year range, GF range
• Click-to-filter tags — click any tag on a card or detail view to filter the dataset
• Cost estimator — quick benchmark-based estimates using filtered comparison sets
• Full-text search across projects, locations, and architects
• Runs entirely in the browser — no server, no build step

Data Sources

Data is extracted from publicly available Bautendokumentationen published by Swiss federal, cantonal, and municipal building authorities.

Source	Organisation	Documents	URL
BBL	Bundesamt für Bauten und Logistik	~144	bbl.admin.ch
armasuisse	armasuisse Immobilien (VBS)	~53	ar.admin.ch
Stadt Zürich	Hochbaudepartement	~36	stadt-zuerich.ch
Stadt Bern	Hochbau Stadt Bern	~55	bern.ch
Stadt St. Gallen	Hochbauamt	~78	stadt.sg.ch
Kanton Aargau	Immobilien Aargau	~7	ag.ch

Extraction Pipeline

The pipeline converts PDF Bautendokumentationen into structured benchmark data in two stages:

Stage 1: PDF to Markdown — Each PDF is converted to clean Markdown using PyMuPDF4LLM (for text-layer PDFs) or Docling (IBM, for scanned/complex-layout PDFs). This preserves table structure — critical for BKP cost breakdowns — and produces human-readable intermediate files.

Stage 2: Markdown to structured data — Markdown tables are parsed to extract BKP/eBKP-H costs, SIA 416 quantities (GF, GV, NGF, HNF), project metadata (architect, client, timeline), and benchmarks (CHF/m², CHF/m³). Multi-language support for German, French, and Italian documents.

The extracted data is loaded into a SQLite database that runs entirely in the browser via sql.js.

See docs/PIPELINE.md for full technical details.

Tech Stack

Frontend

Library	Purpose	License
Vanilla JS (ES6+)	Application logic, no framework	—
CSS Custom Properties	Design token system	—
sql.js	SQLite compiled to WebAssembly, runs in browser	MIT
MapLibre GL JS	Interactive vector map with clustering	BSD-3
Google Material Icons	UI iconography	Apache 2.0

Extraction Pipeline

Library	Purpose	License
PyMuPDF	PDF text and image extraction	AGPL
PyMuPDF4LLM	PDF to Markdown conversion	AGPL
Docling	Advanced PDF parsing with table extraction (IBM)	MIT
Tesseract OCR	OCR for scanned pages (DE/FR/IT)	Apache 2.0
Pillow	Image processing	MIT-like

APIs

Service	Purpose
geo.admin.ch	Geocoding Swiss addresses (free, no key)
Geoapify	Geocoding international addresses
CARTO	Base map tiles via MapLibre

Project Structure

index.html                  Single-page application
css/
  tokens.css                Design tokens (colors, spacing, typography)
  styles.css                Component styles
js/
  db.js                     Database queries
  utils.js                  Shared state, formatting, tag helpers
  views.js                  Gallery, list, map, dashboard renderers
  detail.js                 Detail view, cost tables, carousel, estimator
  main.js                   App init, routing, filters
data/
  kennwerte.db              SQLite database
  pdfs/                     Source PDFs (gitignored)
  markdown/                 Converted markdown (gitignored)
scripts/
  pdf_to_markdown.py        Stage 1: PDF to Markdown
  extract_from_markdown.py  Stage 2: Markdown to structured data
  extract.py                Legacy: direct PDF to DB (v1 regex)
  extract_all.py            Batch wrapper
  download_*.sh/py          Per-source download scripts
docs/
  PIPELINE.md               Extraction pipeline architecture
  DATAMODEL.md              Entity model and field definitions
  SOURCES.md                Data source inventory
  REQUIREMENTS.md           Requirements specification

Documentation

• PIPELINE.md — extraction architecture, tool comparison, processing flow
• DATAMODEL.md — entity model, field definitions, SIA 416 quantities
• SOURCES.md — data source inventory with URLs and status
• REQUIREMENTS.md — functional and non-functional requirements

License

MIT — Digital Real Estate and Support